Re: [Qemu-devel] QEMU event loop optimizations

[Paolo Bonzini]

On 05/04/19 18:33, Sergio Lopez wrote:
>> - qemu_bh_schedule_nested should not be necessary since we have 
>> ctx->notify_me to also avoid the event_notifier_set call.  However, it 
>> is possible to avoid the smp_mb at the beginning of aio_notify, since  
>> atomic_xchg already implies it.  Maybe add a "static void 
>> aio_notify__after_smp_mb"?
> try_poll_mode() is called with ctx->notify_me != 0, so we get at least
> one event_notifier_set() call while working in polling mode.

Right, though then there is the idea of making ctx->notify_me and
ctx->notified bitmasks, where bit 0 is "BH ready" and bit 1 is "poll()
parameters changed".  try_poll_mode() cares about the latter, but does
not need event_notifier_set() for BH ready.

Paolo